Monocular Pedestrian Detection: Survey and Experiments

This article provides an overview of the current state of the art pedestrian detection, a rapidly evolving area in computer vision that has applications in the fields of intelligent vehicles, surveillance, and advanced robotics. The authors first outline and describe the main components of a pedestrian detection system and the underlying models. The second (and larger) part of the article reports on a corresponding experimental study. State-of-the-art systems covered include wavelet-based AdaBoost cascade, HOG/linSVM, NN/LRF and combined shape-texture detection. Experiments are performed on an extensive dataset captured onboard a vehicle driving through an urban environment. The dataset includes many thousands of training samples as well as a 27-minute test sequence involving more than 20000 images with annotated pedestrian locations. The authors consider two evaluation settings: a generic setting, where evaluation is done without scene and processing constraints, and one specific to an application onboard a moving vehicle in traffic. They report a clear advantage of the HOG/linSVM system at higher image resolutions and lower processing speeds, and a superiority of the wavelet-based AdaBoost cascade approach at lower image resolutions and (near) real-time processing speeds. They have made their dataset (8.5GB) public for other researchers for benchmarking purposes.


  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01150428
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Jan 30 2010 11:20AM